Energy-Efficient Self-Organization for Wireless Sensor Networks

Transcription

1 Energy-Efficient Self-Organization for Wireless Sensor Networks Thomas Watteyne CTTC, 22nd May 2007

2 Thomas Watteyne, PhD. candidate CITI Laboratory INRIA / INSA de Lyon France Telecom R&D Grenoble Advisor: Isabelle Augé-Blum Advisor: Mischa Dohler Energy-Efficient Self-Organization / / Thomas Watteyne 2

3 Wireless Sensor Networks Application domains military, surveillance, Health, intelligent homes - measuring a physical value - processing - wireless communication Embedded system processing power memory embedded energy hard to reach area random deployment no human intervention (battery ) no fixed infrastructure changing topology multi-hop Energy-Efficient Self-Organization / / Thomas Watteyne 3

4 Self-Organization: a definition "Self-organization can be defined as the emergence of system-wide adaptive structure and functionality from simple local interactions between individual entities" C. Bettstetter Energy-Efficient Self-Organization / / Thomas Watteyne 4

5 introducing WSNs AnyBody: Self-organization in wireless networks Self-organization in wireless sensor networks Geographic routing with guaranteed delivery 1-hopMAC, an energy-efficient MAC protocol Energy-Efficient Self-Organization / / Thomas Watteyne 5

6 1Introducing Wireless Sensor Networks Energy-Efficient Self-Organization / / Thomas Watteyne 6

7 Sensors Mica2 (2002) Mica2dot (2003) Telos (2004) Smart Dust (???) Cognichip (2007) Energy-Efficient Self-Organization / / Thomas Watteyne 7

8 Applications Natural habitat monitoring Great Duck Island, USA (Berkeley, 2003) Car tracking, sensors deployed from a R/C plane (Berkeley, 2001) Area mapping by flying with a R/C helicopter (Stanford, 2003) Energy-Efficient Self-Organization / / Thomas Watteyne 8 Automated meter reading at Sables-d'Olonne ( 2006)

9 and Networking Application Presentation Session Transport Networking Medium Access Physical layer Find a path between source and destination Medium Access Control Sending the actual data Energy-Efficient Self-Organization / / Thomas Watteyne 9

10 2AnyBody: Self-organization in a Wireless Network Energy-Efficient Self-Organization / / Thomas Watteyne 10

11 Size of a communication range? Energy-Efficient Self-Organization / / Thomas Watteyne 11

12 Traditionally: 1. Neighborhood discovery 2. Cluster forming 3. Interconnecting clusters 4. Hierarchical routing 5. Run-time multi-hop no energy constraints mobility ad-hoc network Energy-Efficient Self-Organization / / Thomas Watteyne 12

13 1. Neighborhood discovery (using HELLO packets) Connectivity graph Energy-Efficient Self-Organization / / Thomas Watteyne 13

14 ? Energy-Efficient Self-Organization / / Thomas Watteyne 14

15 2. grouping nodes into clusters clusterhead 1-cluster Energy-Efficient Self-Organization / / Thomas Watteyne 15

16 2. grouping nodes into clusters 2-cluster Energy-Efficient Self-Organization / / Thomas Watteyne 16

17 2. grouping nodes into clusters Nathalie Mitton, LIFL, France. 1 Energy-Efficient Self-Organization / / Thomas Watteyne 17

18 ? Energy-Efficient Self-Organization / / Thomas Watteyne 18

19 3. interconnecting clusters neighbors gateway Energy-Efficient Self-Organization / / Thomas Watteyne 19

20 3. interconnecting clusters virtual backbone Fabrice Théoleyre, LSR, France. Energy-Efficient Self-Organization / / Thomas Watteyne 20

21 ? Energy-Efficient Self-Organization / / Thomas Watteyne 21

22 4. hierarchical routing protocol D S D 1. S CH S 2. CH S CH X (broadcast) 3. CH D D Energy-Efficient Self-Organization / / Thomas Watteyne 22

23 Energy-Efficient Self-Organization / / Thomas Watteyne 23

24 Cost in energy? 21 nodes Neighborhood discovery: 21 Tx, 44 Rx Clustering and interconnection: 18 Tx, 38 Rx Sending a message: 15 Tx, 35 Rx 54 Tx, 117 Rx Energy-Efficient Self-Organization / / Thomas Watteyne 24

25 Maintaining the structure node mobility / link dynamics nodes appearing / disappearing Maintaining the structure by periodically reconstructing it 21 Tx, 44 Rx per period periodic HELLO packets reconstruction if necessary 18 Tx, 38 Rx per reconstruction Energy-Efficient Self-Organization / / Thomas Watteyne 25

26 5. network run-time any to any communication node mobility supported optimized flooding possible high level protocols such as service discovery can be deployed What are the needs of a Wireless Sensor Network? Energy-Efficient Self-Organization / / Thomas Watteyne 26

27 3Self-organization in a Wireless Sensor Network Energy-Efficient Self-Organization / / Thomas Watteyne 27

28 ad-hoc network vs. WSN no fixed infrastructure ~100 nodes multi-hop no energy constraints bandwitdh needed mobility service deployment sporadic networks no fixed infrastructure nodes multi-hop scarce energy low traffic no mobility application specific long-term networks Energy-Efficient Self-Organization / / Thomas Watteyne 28

29 Self-organization: a reactive approach Traditionally: 1. Neighborhood discovery 2. Cluster forming 3. Interconnecting clusters 4. Hierarchical routing 5. Run-time WSNs: 1. Intelligent routing 2. Run-time Put the intelligence into the routing Energy-Efficient Self-Organization / / Thomas Watteyne 29

30 Routing protocol reactive (avoid maintaining a structure) real-time possible guaranteed message delivery application specific First step: work on geographic routing Energy-Efficient Self-Organization / / Thomas Watteyne 30

31 4Geographic routing with guaranteed delivery Energy-Efficient Self-Organization / / Thomas Watteyne 31

32 Geographic routing: pros and cons Pros Local decisions Simple Guaranteed delivery possible Cons Each node needs to know its position Each node needs to know the destinations position Maintaining a neighborhood table 1hop-MAC current work Energy-Efficient Self-Organization / / Thomas Watteyne 32

33 Greedy geographic routing may fail Energy-Efficient Self-Organization / / Thomas Watteyne 33 If it doesn't fail, near to shortest path

34 Greedy-Face-Greedy Right hand rule to circumnavigate the void planar graph greedy mode? face mode Energy-Efficient Self-Organization / / Thomas Watteyne 34

35 Our proposal Record path in the packet Right hand rule 1. never send a packet to a neighbor whom you have already sent a packet to; 2. send a packet back to a neighbor (i.e. he has sent you a packet before) only if there are no other neighbors you have never communicated with; 3. if you have several choices of neighbors whom you can send back a packet, pick the neighbor who has sent you a packet last. Energy-Efficient Self-Organization / / Thomas Watteyne 35

36 Example greedy mode face mode Energy-Efficient Self-Organization / / Thomas Watteyne 36

37 Results: delivery ratio Energy-Efficient Self-Organization / / Thomas Watteyne 37

38 Results: number of hops ~ same as GFG but we do not assume triangular inequality Can be used when imperfect positioning accuracy Energy-Efficient Self-Organization / / Thomas Watteyne 38

39 GFG fails w. non-perfect positioning When positioning is not perfect, creating a planar graph disconnects the network. Energy-Efficient Self-Organization / / Thomas Watteyne 39

40 51-hopMAC: An Energy-Efficient MAC Protocol for Avoiding 1-hop Neighborhood Knowledge Energy-Efficient Self-Organization / / Thomas Watteyne 40

41 Sources of Energy Wastage Radio budget > 80% total energy budget Collision Overhearing Overheads Idle-listening Layering approach cross-layering thru layer communication or integration Routing geographic based Medium Access Control preamble sampling techniques Physical layer Avoiding neighborhood knowledge Energy efficiency Energy-Efficient Self-Organization / / Thomas Watteyne 41

42 Energy-Efficient MAC coordinated on-off scheduling (synchronization) preamble sampling Micro-frame preamble sampling (A. Bachir et al.) Energy-Efficient Self-Organization / / Thomas Watteyne 42

43 1-hopMAC A f=1 S B f=3 f=2 C S A B C REQ 1. t (f max -f min ). t+tack ACK 3. t 2. t ACK ACK DATA Energy-Efficient Self-Organization / / Thomas Watteyne 43

44 Problem: ACK messages may collide D x 1 =x first x 2 x 3 x N d Energy-Efficient Self-Organization / / Thomas Watteyne 44

45 Calculating the collision probabiliy the larger D, the lower the probability at D=1000, P~10% We're not finished! How to lower it? Energy-Efficient Self-Organization / / Thomas Watteyne 45

46 Reducing collision probability Tune this to have a smaller P using this Metric β (uniformly distributed) mapping function Backoff Time x i (not uniformly distributed) Until now, we considered β and x i to be proportional Energy-Efficient Self-Organization / / Thomas Watteyne 46

47 Result "Nodes should take their metric β, apply the formula above to obtain their x i. This way, the collision probability will be reduced by close to 40%" Energy-Efficient Self-Organization / / Thomas Watteyne 47

48 Energy-Efficient Self-Organization / / Thomas Watteyne 48

49 6Concluding remarks Energy-Efficient Self-Organization / / Thomas Watteyne 49

50 Self-organization in a WSN is not a trivial problem Solutions for ad-hoc network maintain structures By getting the intelligence into the routing layer, we reduce the traffic control We adapt MAC protocol (cross-layering) to be energy efficient Energy-Efficient Self-Organization / / Thomas Watteyne 50

51 References 1. Frey, H. & Stojmenovic, I. On Delivery Guarantees of Face and Combined Greedy-Face Routing Algorithms in Ad Hoc and Sensor Networks. Twelfth ACM Annual International Conference on Mobile Computing and Networking (MOBICOM), Karp, B. & Kung, H. GPSR: Greedy Perimeter Stateless Routing for wireless networks. Annual International Conference on Mobile Computing and Networking (Mobicom), Polastre, J.; Szewczyk, R. & Culler, D. Telos: Enabling Ultra-Low Power Wireless Research. International Conference on Information Processing in Sensor Networks: Special track on Platform Tools and Design Methods for Network Embedded Sensors (IPSN/SPOTS), Prehofer, C. & Bettstetter, C. Self-organization in communication networks: principles and design paradigms. IEEE Communications Magazine, 2005, 43, Younis, O.; Krunz, M. & Ramasubramanian, S. Node Clustering in Wireless Sensor Networks: Recent Developments and Deployment Challenges. IEEE Network, 2006, 20, Energy-Efficient Self-Organization / / Thomas Watteyne 51

52 References 6. Akyildiz, I. F.; Su, W.; Sankarasubramaniam, Y. & Cayirci, E. Wireless sensor networks: a survey. Computer Networks (Elsevier) Journal, 2002, 38, Al-Karaki, J. N. & Kamal, A. E. Routing Techniques in Wireless Sensor Networks: A Survey. IEEE Wireless Communications, 2004, 11, Watteyne, T.; Augé-Blum, I.; Dohler, M. & Barthel, D. AnyBody: a Selforganization Protocol for Body Area Networks. Second International Conference on Body Area Networks (BodyNets), Watteyne, T.; Bachir, A.; Dohler, M.; Barthel, D. & Augé-Blum, I. 1-hopMAC: An Energy-Efficient MAC Protocol for Avoiding 1-hop Neighborhood Knowledge. International Workshop on Wireless Ad-hoc and Sensor Networks (IWWAN), Dohler, M.; Watteyne, T.; Barthel, D.; Valois, F. & Lu, J. Kumar's, Zipf's and Other Laws: How to Structure an Optimum Large-Scale Wireless (Sensor) Network? 13th European Wireless Conference, Energy-Efficient Self-Organization / / Thomas Watteyne 52

53 Thomas Watteyne Energy-Efficient Self-Organization / / Thomas Watteyne 53

54 CC2420 cognichip antenna light sensor slide buttons Energy-Efficient Self-Organization / / Thomas Watteyne 54 push buttons

55 Planar Graph No edges cross. for example using Gabriel Graph Energy-Efficient Self-Organization / / Thomas Watteyne 55

56 Planar Graph Properties no edges cross preserves connectivity (a) Gabriel Graph (b) relative neighborhood graph Energy-Efficient Self-Organization / / Thomas Watteyne 56

57 One of the key issues for a Wireless Sensor Network is energy-efficient selforganization. A Wireless Sensor Network (WSN) consists of a large number of sensors capable of doing three complementary tasks: measuring a physical value, processing that value and communicating over the wireless medium. Most applications imply random deployment of sensors within a to-be-monitored area (think of a forest fire detection network). The nodes in the network send their messages to a specific collecting node called sink node. As their transmission range is small compared to the size of the network, multi-hop communication is needed. A priori, nodes do not know anything about the network they are deployed in. Selforganization is thus needed to created a structure in the network to enable communication. As energy is a major constraint in WSNs, energy-efficiency is vital. The goal of this presentation is to describe different self-organization solutions, and to see how these can be applied to WSNs. We will start by presenting a selection of related works on both wireless sensor networks and self-organization. We will then detail the strong links between energy-efficiency and the different protocol layers. We will finally present two Energy-Efficient Self-Organization / / Thomas Watteyne 57 protocols we have proposed to address the aforementioned constraints.

58 Energy consumption radios on/off Micax family ( x10 4 Telos mote, rev. B a.k.a. t-mote sky ( x16 Telos mote, rev. A ( x10 Energy-Efficient Self-Organization / / Thomas Watteyne 58

59 Berkeley Motes Family Energy-Efficient Self-Organization / / Thomas Watteyne 59

60 Chipcon CC2500 hardware preamble sampling support Energy-Efficient Self-Organization / / Thomas Watteyne 60

61 Right hand rule without planarity E H D G F Energy-Efficient Self-Organization / / Thomas Watteyne 61

62 Right hand rule with planarity E H D G F Energy-Efficient Self-Organization / / Thomas Watteyne 62

63 1-hopMAC var1 S A B C REQ 1. t (f max -f min ). t+tack ACK 3. t 2. t ACK ACK DATA S A B C REQ 1. t (f max -f min ). t+tack ACK 3. t 2. t ACK ACK DATA Energy-Efficient Self-Organization / / Thomas Watteyne 63

64 1-hopMAC var2 S A B C REQ 1. t (f max -f min ). t+tack ACK 3. t 2. t ACK ACK DATA (f max -f min ). t+d+tack S A B C REQ 1. t ACK 3. t 2. t DATA Energy-Efficient Self-Organization / / Thomas Watteyne 64

65 1-hopMAC var3 (f max -f min ). t+d+tack S A B C REQ 1. t ACK 3. t 2. t DATA (f max -f min ). t+d+tack S A B C REQ 1. t ACK 3. t 2. t DATA Energy-Efficient Self-Organization / / Thomas Watteyne 65

66 Analysis extract the radio time of each variant compare them two-by-two 1. 1-hopMAC var1 always better than 1-hopMAC basic 2. 1-hopMAC var2 better than 1-hopMAC var1 iff first ACK received before t thresh = f max. t+(2-n)t ACK +2d 3. 1-hopMAC var3 always better than 1-hopMAC var2 1-hopMAC var3 if first ACK before t thresh 1-hopMAC var1 if first ACK after t thresh Energy-Efficient Self-Organization / / Thomas Watteyne 66